drm/amdgpu: add option to clear VM page tables after every submit

This makes it much easier to find when userspace misses to send some buffers.
Signed-off-by: NChristian König <christian.koenig@amd.com>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -80,6 +80,7 @@ extern int amdgpu_deep_color;
 extern int amdgpu_vm_size;
 extern int amdgpu_vm_block_size;
 extern int amdgpu_vm_fault_stop;
+extern int amdgpu_vm_debug;
 extern int amdgpu_enable_scheduler;
 extern int amdgpu_sched_jobs;
 extern int amdgpu_sched_hw_submission;

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

@@ -567,9 +567,24 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p,
 			if (r)
 				return r;
 		}
+
+	}
+
+	r = amdgpu_vm_clear_invalids(adev, vm, &p->ibs[0].sync);
+
+	if (amdgpu_vm_debug && p->bo_list) {
+		/* Invalidate all BOs to test for userspace bugs */
+		for (i = 0; i < p->bo_list->num_entries; i++) {
+			/* ignore duplicates */
+			bo = p->bo_list->array[i].robj;
+			if (!bo)
+				continue;
+
+			amdgpu_vm_bo_invalidate(adev, bo);
+		}
 	}
 
-	return amdgpu_vm_clear_invalids(adev, vm, &p->ibs[0].sync);
+	return r;
 }
 
 static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev,

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -76,6 +76,7 @@ int amdgpu_deep_color = 0;
 int amdgpu_vm_size = 8;
 int amdgpu_vm_block_size = -1;
 int amdgpu_vm_fault_stop = 0;
+int amdgpu_vm_debug = 0;
 int amdgpu_exp_hw_support = 0;
 int amdgpu_enable_scheduler = 1;
 int amdgpu_sched_jobs = 16;
@@ -145,6 +146,9 @@ module_param_named(vm_block_size, amdgpu_vm_block_size, int, 0444);
 MODULE_PARM_DESC(vm_fault_stop, "Stop on VM fault (0 = never (default), 1 = print first, 2 = always)");
 module_param_named(vm_fault_stop, amdgpu_vm_fault_stop, int, 0444);
 
+MODULE_PARM_DESC(vm_debug, "Debug VM handling (0 = disabled (default), 1 = enabled)");
+module_param_named(vm_debug, amdgpu_vm_debug, int, 0644);
+
 MODULE_PARM_DESC(exp_hw_support, "experimental hw support (1 = enable, 0 = disable (default))");
 module_param_named(exp_hw_support, amdgpu_exp_hw_support, int, 0444);